A Logit Model to Predict Bond Ratings in India

Abstract

In this paper, the authors have attempted to estimate a logit model to predict bond ratings. The prediction of bond rating assumes significance because investors can use this information to adjust their portfolios to get better results. Also, issuing companies can also utilise this information to reconsider and realign their capital budgeting and investment policies. The financial data on issuing companies used in the study, was taken from the website “www.indiainfoline.com”. Various financial parameters and ratios of 40 companies were considered to estimate a logit model. The companies having rating above LBB+ (ICRA rating for long-term debt) were considered as having a good rating and the ones having rating LBB+ and below were considered as having a poor rating. The model correctly classified the bond ratings of 95 per cent of the sample companies. The model was also used to predict the ratings of 10 holdout companies and it was found that the ratings of 8 out of 10 companies were correctly predicted. Two variables, namely, interest cover and net sales to total assets, were found to be significant. It is suggested that the model building exercise be taken up on a continuous basis as the significance of variables may undergo a change over time.